Traditional path planning methods, such as sampling-based algorithms, struggle with high computational complexity in high-dimensional spaces like those encountered with 6-DOF robotic arms. Deep reinforcement learning offers a promising alternative, but its efficiency is often hindered by sparse rewards and large action spaces. To address these challenges, we integrate human demonstrations to guide the learning process, thereby improving the convergence rate and policy performance of the SAC algorithm. This paper proposes a novel approach for robotic arm path planning in complex product assembly scenarios using a soft actor-critic (SAC) deep reinforcement learning algorithm augmented with human demonstrations. The proposed method involves designing a specialized state space and reward function based on key points provided by human demonstration. The proposed approach is validated through simulations in a complex environment with obstacles, demonstrating significant improvements in planning efficiency and adaptability. The results show that incorporating human demonstrations effectively enhances the robot's ability to learn optimal collision-free paths, making it a viable solution for the path planning of complex assembly tasks.